Well faster and faster... Depends on what offcourse a graphics card is faster at scanline rendering. However its not getenrally programmable in same way. So no its not faster on mentalray atleast not untill mentalray is ported to cuda. But see the gfx card is much more limited in genral ability tough at that MUCH faster because of this*
A tesla is the same thing your normal higher end quadros are except it has several quadrons in the same rack. Its designed for computation centers and scientific computation for those who actually have a server room who arent over porting code form a system to another for speed. Essentially its a computer that ONLY has a graphics card, since its not gfx card per see you cant easily plug it to your average computer, you need a disk server and networks to do the communication.
Its not targeted for you and you can be safe you wont be running anything most people would, on it.
*its a sacrafice if you dedicate something on the cpu for something it is up to 1000 time faster but is not as flexible its a trade off.